NVIDIA Datacentre GPUs
Powering the ultimate servers for delivering professional graphics
How to Choose the Perfect Datacentre GPU
The GPU (Graphics Processing Unit) accelerator or graphics card is the most important component in a server designed to deliver a professional graphics platform from a centralised datacentre location. Datacentre GPUs are passively cooled cards designed for 19in rack servers, providing high performance compute for workloads such as deep learning and AI, HPC, rendering and simulations. The choice of GPU will determine the overall performance and perhaps more importantly the number of concurrent users that can be supported without loss of integrity in the graphics quality. The server could be supporting shared graphics for a workgroup or delivering a larger virtual service for an entire organisation such as NVIDIA Virtual PC (vPC), Virtual Workstation (vWS) or Omniverse. NVIDIA professional GPUs are specifically designed to cope with these high-performance workloads and have several attributes that set them apart from their consumer counterparts.
Certified Drivers
Enterprise Class
ECC Memory
Extended Memory
Security
Extended Warranty
With prices ranging from over around £2,000 up to over £10,000, there’s an NVIDIA datacentre GPU for every project and budget. The range is made up the latest Ampere-based cards, and their use case suitability is summarised in the table below.
| A100 | A30 | A40 | A10 | A16 | A2 | |
|---|---|---|---|---|---|---|
| Deep Learning Training | Yes |
Yes |
No |
No |
No |
No |
| Deep Learning Inferancing | Yes |
Yes |
No |
Yes |
No |
Yes |
| HPC | Yes |
Yes |
No |
No |
No |
No |
| Rendering | No |
No |
Yes |
Yes |
No |
No |
| Cloud Workstation | No |
No |
Yes |
Yes |
No |
No |
| Cloud PC | No |
No |
No |
No |
Yes |
Yes |
NVIDIA Datacentre GPU Accelerators
To help you choose between the different GPUs the summary table below rates the various performance attributes out of 10. The table also indicates which of the datacentre GPUs support ray tracing and are compatible with VR applications.
| A100 | A30 | A40 | A10 | A16 | A2 | |
|---|---|---|---|---|---|---|
| Visualisation Performance | N/A | N/A | 10 (vGPU only) | 9 (vGPU only) | 3 (vGPU only) | 3 (vGPU only) |
| Compute Performance (FP64) | 10 | 6 | 3 | 3 | N/A | 1 |
| Compute Performance (TF32) | 10 | 7 | 7 | 6 | N/A | 2 |
| Compute Performance (FP32) | 6 | 5 | 10 | 8 | N/A | 3 |
| Compute Performance (TF16) | 10 | 7 | 7 | 6 | N/A | 3 |
| Compute Performance (FP16) | 10 | 7 | 7 | 7 | N/A | 3 |
| Compute Performance (INT8) | 10 | 7 | 7 | 3 | N/A | 3 |
| Ray Tracing | No |
No |
Yes (vGPU only) |
Yes (vGPU only) |
Yes (vGPU only) |
No |
| VR Ready | No |
No |
Yes (vGPU only) |
Yes (vGPU only) |
Yes (vGPU only) |
No |
| PRICE |
|
|
|
Call for price |
|
Call for price |
| VIEW > | VIEW > | VIEW > | VIEW > | VIEW > | VIEW > |
